Sampling Methods. Tasks Population vs. Sample Population vs. Sample Sources of sampling error Sources of sampling error Sample size & response rates Sample.

Sampling Methods

Tasks

Population vs. Sample

Sources of sampling error

Sample size & response rates

Probability sampling & its methods

Non-probability sampling & its methods

What is sampling?

Process of choosing subjects for inclusion in the study

Individuals, teams, groups, agencies, sports

Called - subjects/participants

Choosing a Sample

Population Total group to which results

can be generalized

NFL sales staff DI Tennis Players 5th grade girls Male intramural players

N=2,554

Sampling Frame All the people who

compose a particular group

AFC sales staff MVC tennis players 5th grade girls in 2 districts Male intramural players in

schools 20-40,000 pop

n=2,554

Choosing a Sample

Population Delimiting variables Sample

Delimiting variables:• Demographic variables that narrow population

• Age, gender, geography

• Other variables• Group, division, sport, sector, grade

Choosing a Sample How does alcohol sales within a collegiate

stadium impact ticket sales?

What motivates runners to participate in “fun” runs such as color runs, zombie runs, etc?

Sources Error

2 Sources of Error

1. Sampling error

2. Non-sampling error

Sampling Error The difference between characteristics of a sample

& the characteristics of the population

Get a representative sample

The smaller the error, the more reliable the data

As sample size increases, error rates decrease

Sampling Error Error rates

Calculated statistically

50% with + 4 points = 46% - 54%

Political polls…

Sampling Error

Pop. Obama Romney Undecided Margin Error

Rasmussen1,500 LV 48 46 - Obama +2 ±3.0

Ipsos/Reuters

855 LV 47 42 8 Obama +5

Gallup

3,050 RV 50 44 - Obama +6 ±2.0

LV=Likely votersRV=Registered voters

56.8% of registered voters vote in presidential election

Sampling Error

Who responds… Better educated Higher socioeconomically Higher need for social approval More sociable Somewhat less conventional Less conforming Female

Non-Sampling Error

Biases that exist due to who answers a survey Question confusion…validity

Accessibility questions Confusion on terms

Lack of knowledge by respondent Don’t answer vs. Neutral response

Concealment of the truth

Non-Sampling Error

Biases that exist due to who answers a survey Loaded questions Don’t you agree that social workers should earn

more money than they currently earn? ___ Yes, they should earn more ___ No, they should not earn more ___ Don’t know/no opinion

Do you believe social worker salaries are a little lower than they should be, a little higher than they should be, or about right?

Non-Sampling Error

Biases that exist due to who answers a survey Weighted scales

Examples…

Excellent Good Fair Poor

Strongly agree Agree Disagree Strongly disagree

Non-Sampling Error

Examples Asking college age students about family finances

Surveying those completing class & not those registered & dropped it

Pontiac Parks & Rec - Surveying those who are members

Sample Size

Goal Collect a sample that is large enough to be

representative of the population, but not so large as to waste resources

Determining sample size Use population if it is small Literature Statistics to run… X number needed for analysis Table…

Sample Size

Notes:

5% = 5% chance the sample differs from the population

point of diminishing returns

Response Rates

# within your sample who complete the survey 70% special interest groups

Parents, fans 60% professional groups

Staff & Volunteers 55% general interest

Response Rates

Sample size vs response rates Estimate # needed based on sample size chart Estimate response rate Inflate sample size to accommodate for

nonresponses

Will get some unusable surveys & addresses

WHAT INCREASES YOUR RESPONSE RATES?

Increase Response Rates

Pre-notification Contact respondents in advance Give opt out option

Interest in topic

Survey design Short & concise

Increase Response Rates

Timing & delivery Holidays, pool openings, summer vacation, NCAA

tournament

Incentives Younger audience – electronics Older audience – gift cards, free conference reg.

Send reminders

Increase Response Rates

E-mail invites Professionals

avoid Friday-Monday Students

Monday afternoon, Thursday morning, Saturday afternoon

Avoid spam language Personalize the e-mail with respondents name Use clean, updated list

Sampling Methods

1. Probability Sampling Simple Random Sampling Stratified Random Sampling Systematic Sampling Cluster Sampling

2. Non-probability Sampling Purposive Sampling Convenience Sampling Quota Sampling Snowball Sampling

All members of the population have a chance of being selected

Sample is not drawn by chance

Simple Random Sampling

Equal probability of being selected

Results in the most reliable data

Will most represent the population

How to do it: Draw names, teams, leagues, classes Assign numbers Random numbers table… Software

1. Needs: 5 random numbers between 0-20

2. Randomly select a row.

3. Read 2 #’s at a time, select those that are between 00-20.

Simple Random Sampling

Software

http://www.randomizer.org/lesson4.htm

Simple Random Sampling

Strengths No subject classification error Easy to understand

Weaknesses Have to number each person Larger sampling error than stratified random

sampling

Stratified Random Sample Randomly selected from within a stata

(subpopulation)

Need to be able to assign everyone to one strata

Age, race, gender, income, geography

Allows researcher to compare groups

Stratified Random Sample Non-proportional sampling

# selected from each strata Gold Medal finalist/winning agency directors

Proportional sampling Class make-up = 60% boys; 40% girls

Sample = 60% boys; 40% girls

Example Male vs. female

Female: 22/41 = 54% Male: 19/41 = 46%

Sample size = 54% x 36 = 19 Females 46% x 36 = 17 Males

Kettering phone survey

Population

Sample

Stratified Random Sample Strengths

Can compare subgroups More representative than simple random

sampling Results represent population

Weaknesses Requires subgroup classification Need to know the proportion of each group Costly

Systematic Sampling Determine a rationale for a sampling routine

Select every “nth ” person

5,000 population, 370 sample Every 10th person starting with 5th person Roll the dice; draw a number

ISU Women’s b-ball attendees Every 4th person to pass by

Systematic Sampling Strengths

Simple process Don’t have to classify or number people

Weaknesses Larger sampling error than Stratified RS

Cluster Sampling Divides population into naturally occurring groups or

units rather than individuals Neighborhoods Conferences Grades

Use specific units to randomly select or stratify NASPD Regions – randomly select 4 of the 8 regions ROE – 1 of 3 counties, schools within

Cluster Sampling Strengths

Low cost Can analyze individual clusters

Weaknesses Higher error than simple random & stratified random Requires everyone assigned to 1 cluster

Probability Sampling Overview Least Error….

Stratified random sampling Simple random / Systematic Cluster

Non-probability Sampling

Used when population is unknown Fans People with a specific disability Runners, bikers, hikers, backpackers

Sample isn’t drawn by chancePurposive SamplingConvenience SamplingQuota SamplingSnowball Sampling

Purposeful/Purposive Sampling Select certain individuals because you feel

they represent the entire population Groups, classes/programs, time of day MVC Campus Rec Departments

Qualitative Select “info rich” cases Key informants Few will give in depth knowledge Generalization isn’t the purpose

Purposeful/Purposive Sampling Strengths

Easy to administer Less costly & time consuming Generalization possible to similar subjects

Weaknesses Difficult to generalize to other subjects Experimenter subject bias

Convenience Sampling Chosen because they are accessible

Ie. Survey my classes, ISU students, IWU S-A’s

Higher error rates, less genralizability

ISU Volleyball non-attendees Survey classmates Dorm dwellers

Convenience Sampling Strengths

High participation rates Cheap, easy Generalization to similar subjects

Weaknesses Difficult to generalize to other subjects Experimenter subject bias

Quota Sampling Divide population into sub-groups

Survey equal number of each group Stratified is a % Cluster is a section Quota is = numbers

May not be representative of the population

Quota Sampling Strengths

High participation rates Cheap, easy Generalization to similar subjects More representative sample

Weaknesses Same as others More time consuming than others

ComparisonN=500 alums n= 217

Sport Mgt Recreation PETE

250 (50%) 150 (30%) 100 (20%)

* Stratified Random (proportional)

109 65 43

* Cluster Sampling Select this group only

- -

** Quota Sampling* Stratified Random (non-proportional)

73 73 73

• * Probability• ** Non-Probability

Snowball Sampling

Based on recommendations

Stay at home dads

Women motorcyclists

Athletes recruited by but not attending ISU/IWU

Weaknesses of NPS

Generalizability Used most if purpose is to understand & not to

generalize Show how the sample matches the population Indicate that results will be same for the sample

population Biased sample

Bias by whom you select